Professional Documents
Culture Documents
223_CP2_P1_Audit trail
223_CP2_P1_Audit trail
223_CP2_P1_Audit trail
EXAMINATION
Sept 2023
AUDIT TRAIL
Paper One
This is done by analysing the temperature changes over the past years from the given data, fitting
a linear regression model to the data and hence predicting future temperatures by identifying a
trend. This will help assess how future refrigeration costs are expected to change with the
predicted changes in temperature.
Data
Data was provided by my manager. (The source is the National Centres for Environmental
Information (NCEI)). It consists of two tables: The Stations tab holds information on each
weather station our client has operations in (location, elevation and country/continent). The Data
tab holds average monthly temperature in degrees celsius for each year for each weather station.
The years range from 1961 to 2010.
The calculations and checks are copied as well, and the temperature graph shows more
reasonable values for minimum temperatures - again, this is affected by the bias towards the
Northern hemisphere, and shows a marked increase in summer and a decrease in winter.
Assumptions
The following assumptions have been made:
• After the adjustments described above, the data is fit for purpose and accurate.
• Summer months apply consistently as April to September across the whole northern
hemisphere, irrespective of latitude
• No more recent suitable data is available after 2010
• No event or other changes have taken place to invalidate the trend obtained from the data
up to 2010.
• No change in business conditions that affect refrigeration costs will take place between
2020 and 2030
Inputs
This worksheet provides a list of all parameters to support the calculations. This includes inputs
to split locations by latitude above and below a certain point (set to zero), and by elevation (set to
500).
Also here is information about costs in 2020 and the constant parameter for the formula used to
determine the impact of temperature on refrigeration costs.
Method
Each sheet is dealt with in turn.
Data Prep
This sheet does the analysis of the data and prepares the summary from which the later graph and
linear regression will be done.
Column P contains the Hemisphere category - whether this is North (latitude >0) or South
(latitude < 0) based on a lookup of the weather station ID on the Stations Adj tab.
Column Q contains the elevation category - whether the station is above or below the input
elevation parameter of 500m. The result is returned as High or Low based on a lookup on the
Stations Adj tab.
Column R contains the average temperature across the year - taking the average of columns C to
N.
Column S contains the average temperature for Winter. This is the average of April to September
for Southern Hemisphere stations (based on column P), and the average of the other months (Jan
to March and October to December) for Northern Hemisphere stations.
Column T contains the average for Summer - this is the same as for Winter, but reversed.
Column U picks up the continent based on a lookup of Weather station ID from the Stations Adj
tab
Analysis
This tab calculates average temperatures for each year for each category.
The Averageifs formulas is used throughout to calculate the average across all data points that
meet the provided criteria. Unless otherwise specified, the average of column R on the Data Prep
sheet is taken.
Column B calculates the overall average temperature across all stations for each year.
Column C and D calculate the average temperature across all Northern Hemisphere and Southern
Hemisphere stations respectively for each year (using the category in column P of the Data Prep
sheet)
Column E and F calculate the average temperature for all Low and High elevation stations
respectively for each year (using the category in column Q of the Data Prep tab).
Column G and H calculate the average Winter and Summer temperature for each year (taking the
average from column S and T on the Data Prep tab respectively).
Columns I to N calculate the average temperature for each continent each year (using the
continent in the Data Prep tab column U).
Charts
This tab contains the following four line charts showing movement of temperature over the years
from 1961 to 2010, broken down as follows:
• Weather stations north and south of the equator, as well as an overall average. This also
shows the bias to stations in the north: the overall average is much closer to the north
line. The north average is much lower than the south, and shows slightly more volatility
(such as from 1995-1999).
• Weather stations higher and lower than 500m elevation. This has a minimal impact on
average temperature, with the three lines very closely banded. While higher elevation
should (and does) have lower average temperatures, it is likely that most weather stations
would not be at the really high altitudes which would have a more dramatic impact on
temperature.
• Average temperature for summer and winter. Summer is much higher than winter, as
expected. Summer temperatures also appear to be smoother over time than winter.
• Average temperature by continent. Africa and South America are well clear above 20°,
with Europe notably lower than the rest below 10°. It appears that the continents with
higher temperature have lower volatility.
Model
This tab uses a linear regression model to predict future summer temperatures, and uses those to
gauge the impact on refrigeration cost.
At the top of the sheet, in columns B and C, the LINEST formula is used to do a linear
regression. This formula is an array formula, so care must be taken when editing this to select
both cells B6 and C6, and use Ctrl-Shift-Enter to enter it. The formula takes Y values as the
average summer temperatures from column H in the Analysis tab, and X values as the year.
Below this, the actual and predicted summer temperatures are calculated. From 1961 to 2010 the
formula just links to column H on the Analysis tab. From 2011, the formula above is used to
calculate the predicted temperatures. The cell for 2011 is highlighted to avoid formulas being
copied down.
In columns F-H, the average temperature for 2006-2010 (in H13) and 2026-2030 (in H14) are
calculated.
The average costs for 2026-2030 are also calculated just below this.